Deirdre Bosa - CNBC thumbnail

Deirdre Bosa - CNBC

Perplexity CEO Aravind Srinivas on Comet, AI Browser Wars, and Why He’s Not Selling to Big Tech

🎥 Perplexity CEO Aravind Srinivas on Comet, AI Browser Wars, and Why He’s Not Selling to Big Tech

⏱️ Duration: 25:09
🔗 Watch on YouTube

Overview

This video features an in-depth interview with Aravind Srinivas, CEO of
Perplexity, discussing the launch and early feedback of their new AI-powered
browser, Comet. The conversation explores Comet’s capabilities, user adoption,
privacy considerations, the evolving AI landscape, and Perplexity’s long-term
vision to create a truly agentic browser that acts as a digital Swiss Army knife
for productivity.


Main Topics Covered

  • Launch and Early Adoption of Comet Browser
  • User Feedback and Unexpected Use Cases
  • Comet’s Vision: An AI-Powered, Agentic Browser
  • Privacy, Security, and Data Handling
  • Challenges of Mainstream Adoption and Scaling
  • Business Models and Monetization Strategies
  • Competition and Differentiation in the AI Landscape
  • Open Source AI Models and Global AI Dynamics
  • Talent and Hiring in the AI Industry
  • Technical Roadmap: Local Inference and Device Integration
  • Future Directions: Beyond the Browser

Key Takeaways & Insights

  • Comet’s Unique Agentic Approach: Unlike traditional chatbots, Comet enables end-to-end workflows by letting users delegate complex tasks to an AI agent that can interact with web pages, control browser tabs, and automate repetitive actions.
  • Unexpected User Creativity: Early users have leveraged Comet for diverse workflows, such as home automation, customer support, market research, unsubscribing from emails, and price comparison for small businesses.
  • Browser as an Operating System: Perplexity aims to turn the browser into a central hub for personal and business productivity, akin to an AI operating system that manages tasks, context, and data securely.
  • Privacy by Design: User data remains on the client side, with Comet accessing third-party services only as needed, and without storing personal data on Perplexity’s servers—enhancing privacy compared to desktop apps.
  • Scaling and Monetization Challenges: The compute-intensive nature of agentic AI makes free scaling difficult; Perplexity is exploring usage-based pricing and premium tiers for advanced features while keeping basic functionality free.
  • Independence From Big Tech: Perplexity sees value in remaining independent and offering an alternative to dominant platforms like Google, resisting acquisition and focusing on execution and retention.
  • Open Source and AI Democratization: While US and global trends in open source AI are in flux, Perplexity embraces open models for post-training and cost reduction, and sees international innovation as healthy competition.
  • Hiring and Talent Development: Instead of competing for top AI researchers with massive salaries, Perplexity focuses on developing raw talent by giving engineers ambitious, differentiated problems to solve.

Actionable Strategies

  • Adopt Agentic Workflows: Explore delegating complex or repetitive digital tasks—such as summarizing content, managing communications, or automating research—to Comet’s AI agent.
  • Leverage AI for Productivity: Use Comet to streamline processes like comparing prices, crafting marketing messages, or managing information across multiple platforms, reducing manual work.
  • Control Data Access: Take advantage of Comet’s privacy features by using incognito mode or limiting which AI features can access your data, depending on your comfort level.
  • Prepare for Usage-Based Pricing: Consider which AI-powered workflows add the most value, as advanced features may shift to a pay-per-use or subscription model.

Specific Details & Examples

  • User Use Cases: Examples include using Comet to control home devices (e.g., changing light color based on weather), conduct package tracking via customer service, automate email unsubscribing, and monitor news or stocks.
  • Business Applications: Small businesses use Comet for price comparison, social media research, and drafting marketing messages, saving significant time and money (e.g., reducing two hours of price comparisons to five minutes, saving $200).
  • Waitlist and Demand: Within five days of launch, Comet’s waitlist exceeded half a million users.
  • Model Architecture: Comet leverages a mix of Perplexity’s own models, DeepSeek, OpenAI, and Anthropic models; it is model-agnostic, choosing the best tool for each task.
  • Privacy Mechanism: Comet accesses third-party services only through browser tabs, using existing login sessions and cookies, and never stores passwords or long-term personal data.
  • Scaling Infrastructure: To support increased usage, Perplexity is rebuilding its search infrastructure and optimizing for high retention before raising additional capital.
  • Open Source Models: References to Moonshot’s Kimi model (comparable to Claude Opus) and the desire to run inference locally on user devices, especially Macs with powerful hardware.

Warnings & Common Mistakes

  • Over-Permissioning Concerns: Users may be wary of granting Comet access to their data; however, the design ensures data stays client-side unless explicitly used for a task.
  • Adoption Curve: There is a learning curve for users to fully utilize agentic workflows, as this mode of interaction differs from instant-response chatbots.
  • Compute Costs: Scaling agentic AI is expensive—users and developers should be aware that not all features can remain free indefinitely.
  • Privacy Misunderstandings: Users should understand the distinction between client-side data usage and server-side data storage, especially compared to traditional desktop AI apps.

Resources & Next Steps

  • Comet Browser Waitlist: Interested users can join the waitlist to try Comet as it rolls out.
  • Basic AI Features Remain Free: Summarization and simple AI search will remain free, while advanced, agentic and productivity features may adopt usage-based or subscription pricing.
  • Stay Informed on Open Source: Watch for developments in open source AI models, as Perplexity intends to incorporate and post-train on the best available models.
  • Feedback and Experimentation: Users are encouraged to experiment with Comet’s capabilities and provide feedback, informing future feature development and mainstream adoption strategies.
  • Business and Technical Partnerships: Perplexity remains open to collaborations, especially with publishers, and is exploring integrations beyond the browser into desktop environments.

This summary encapsulates the video’s detailed discussion on the launch, current
capabilities, user adoption, privacy, competition, and future direction of
Perplexity’s Comet browser, highlighting its ambition to redefine productivity
and AI-powered digital interaction.

← Back to Deirdre Bosa - CNBC Blog